Pemodelan Smart Agriculture untuk Perencanaan Sistem Pemeliharaan Tanaman

Abstract

The planning model for sustainable crop maintenance associated with smart agriculture is a complicated issue because it involves many factors such as productivity, quality, growth care, labor use, and the use of information technology. This study produced an optimization model of Convolutional Neural Network (CNN) that can classify and monitor diseases in plants. The authors propose a new model called Grouping Uniform Neural Network or GUNNet. This model is formed by modifying the ResNet model by reducing the number of layers and changing the pooling type and grouping datasets into uniform sizes. Researchers created a simpler but still effective version for a few datasets. Researchers modified ResNet with depthwise separable convolution and used average pooling as an alternative to max pooling. The model can adaptively select the architecture that matches the number of datasets to be trained. The proposed model is divided into two classifications, namely healthy plants, plants attacked by diseases in the form of insect pests and plants attacked by fungi. From the results of this study can recommend and contribute to the planning and maintenance of plants.